Optimization of Photomixers and Antennas for Continuous-Wave Terahertz Emission

IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 41, NO. 5, MAY 2005 717

Optimization of Photomixers and Antennas forContinuous-Wave Terahertz Emission

Ian S. Gregory, Colin Baker, William R. Tribe, Ian V. Bradley, Michael J. Evans, Edmund H. Linfield,A. Giles Davies, and Mohamed Missous, Member, IEEE

Abstract—We have studied terahertz emission from interdigi-tated finger photomixers coupled to planar antenna structures.Using both pulsed and continuous-wave excitation, polarizationmeasurements reveal that the antenna design dominates the prop-erties of the radiated output at frequencies below 0.6 THz, whilethe efficiency at higher frequencies is additionally dependent onthe design of the photomixer fingers. We have produced terahertzmaps of the device, characterizing the photomixer by measuringthe generated power as a function of the excitation position. To-gether, these measurements have allowed us to understand betterthe distinct roles of the photomixer and antenna in emission atdifferent frequencies and, hence, independently optimize thesecomponents.

Index Terms—Continuous-wave (CW), LT-GaAs, photoconduc-tive antenna, photomixer, terahertz.

I. INTRODUCTION

THE terahertz region of the electromagnetic spectrumranges from 100 GHz to 10 THz, between the millimeter

and the far-infrared frequencies. Compared to many otherregions of the spectrum, scientific and technological progressat terahertz frequencies has been largely impeded by a lackof coherent sources and detectors. Although such devices arecommonplace at the infrared and microwave extremes of theterahertz range, they cannot be easily modified to work atterahertz frequencies. Recent developments have started toaddress this issue, and a variety of novel emitters and detectorsbased on semiconductor technology are emerging [1]–[4].Terahertz radiation is nonionizing, has a shorter wavelength,and hence higher resolution, than microwaves, and is minimallyattenuated by many common materials. Consequently, the tech-nological advances have been accompanied by much interest in

Manuscript received July 19, 2004; revised September 21, 2004. This workwas supported in part by an EC-Framework V program (WANTED), theEPSRC, the Royal Society (A. G. Davies) and in part by Toshiba ResearchEurope Ltd. (E. H. Linfield). Aspects of this work are contained in U.K. PatentApplication 0221056.5.

I. S. Gregory is with the Semiconductor Physics Group, Cavendish Labora-tory, University of Cambridge, Cambridge CB3 0HE, U.K., and also with Ter-aView Ltd., Cambridge CB4 0WS, U.K. (e-mail: [email protected]).

C. Baker, W. R. Tribe, I. V. Bradley, and M. J. Evans are with TeraView Ltd.,Cambridge CB4 0WG, U.K. (e-mail: [email protected]; [email protected]; [email protected]; [email protected]).

E. H. Linfield and A. G. Davies are with the School of Electronic andElectrical Engineering, University of Leeds, Leeds LS2 9JT, U.K. (e-mail:[email protected]; [email protected]).

M. Missous is with the Department of Electrical Engineering and Electronics,University of Manchester Institute of Science and Technology, Manchester M601QD, U.K. (e-mail: [email protected]).

Digital Object Identifier 10.1109/JQE.2005.844471

Fig. 1. Schematic diagram of the photoconductive emitter and identicaldetector, showing the metallic planar antennas, illuminated photoconductiveswitches, and the silicon lenses. The same arrangement is employed for bothpulsed and CW-THz implementations.

possible applications in fields including security screening [5],spectroscopy, and medical imaging [6].

Established experimental terahertz systems are largely basedon photoconductive switches, which rely on the production offew-cycle terahertz pulses using an ultrafast (femtosecond) laserto excite a photoconductive antenna. This technique is inher-ently broadband, with the emitted power distributed over sev-eral terahertz. In the emitter, a transient change in the resistanceof a photoconductive switch occurs on a terahertz timescalewhen illuminated by a laser pulse. Application of an external dcbias across the switch creates a current flow that contains com-ponents at terahertz frequencies. These currents induce a tera-hertz electromagnetic field in the planar, metallic antenna con-nected to the switch. The resulting terahertz dipole radiation iscoupled from the antenna into free space as a quasi-collimatedbeam using a high-resistivity hyper-hemispherical silicon lens[7]. Optoelectronic detection is possible by measuring the cur-rent induced in a similar antenna circuit by the incoming ter-ahertz radiation [8]–[11]. By gating the receiver switch with asecond femtosecond pulse synchronized to the terahertz emis-sion, a dc signal may be measured. Varying the optical pathlength to the receiver allows the entire terahertz time domain tobe sampled. Hence both the amplitude and phase of the incidentterahertz wave can be obtained, and a dynamic range of 60 dBdemonstrated using time-gated detection [12]. Fig. 1 shows aschematic diagram of the photoconductive switch and antenna,

0018-9197/$20.00 © 2005 IEEE

718 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 41, NO. 5, MAY 2005

Fig. 2. Schematic diagram of a CW-THz device, consisting of an antennacoupled to the external circuit. Inset are two designs for electrodes: (a) simplephotoconductive gap and (b) with interdigitated fingers attached. The structureshown in the inset, comprising the photoconductor and electrodes, is generallytermed the “photomixer.”

along with a typical free-space terahertz system incorporatingsilicon lenses.

Terahertz-pulsed imaging has been widely reported [6], [13],[14], with image capture by serial pixel collection. Since in-formation about the entire waveform is available at each pixel,a spectroscopic analysis is simultaneously possible. However,the commercial application of such systems has highlighted anumber of limitations. Primarily, the femtosecond laser and as-sociated chillers and power supplies are bulky and expensive,and thus not ideal for portable terahertz systems for use outsideof the research laboratory.

An alternative approach to terahertz systems is optical hetero-dyne conversion, or photomixing, which can be achieved usingtwo continuous-wave (CW) lasers [15], [16]. The mixing of twoabove-bandgap (visible or near-infrared) wavelengths producesbeating, which can modulate the conductance of a photocon-ductive switch (semiconductor) at the terahertz difference fre-quency. Upon application of a bias, monochromatic CW-tera-hertz (CW-THz) radiation is produced. Furthermore, Vergheseet al. have demonstrated that coherent homodyne detection ispossible in the reverse scheme [17], in analogy with the opticalgating mechanism for photoconductive detection of pulsed tera-hertz radiation. Since these all-photoconductive systems can bedriven by inexpensive, compact and tunable diode lasers [18],[19] they may become a cost-effective commercial terahertzproduct.

The design of CW-THz emitters and detectors consists of sev-eral distinct components. These are the photoconductor, specif-ically the semiconductor material, the electrodes, which definethe geometrical arrangement of the photoconductive gap, andthe antenna. A schematic of a typical CW-THz device is shownin Fig. 2. The performance of such a device can be understoodand optimized by considering separately the roles of the var-ious components, and applying an equivalent electrical circuitmodel to analyze their interconnection. The role of the photo-conductor is to provide a modulated conductivity in responseto the optical field, through the photo-creation of electron–holepairs. The electrodes couple the charge within the photocon-ductor to the antenna and external circuit, and their design isused to optimize the efficiency of the optical modulation. Theantenna is designed to optimize the coupling between the cir-cuit and free-space electromagnetic radiation modes.

Fig. 3. Equivalent circuit diagram for the photomixer and antenna. Currentdrawn from the bias source is modulated at angular frequency ! by thephotoconductor, with resistance R(!; t). The capacitance C represents theeffects of the charge accumulating at the electrodes. Power is dissipated in theantenna, R (!), of which the component oscillating with angular frequency! is coupled out as terahertz radiation.

The antenna and electrodes are, of course, a connectedmetallic structure, and are bonded to the external circuit. Foremitters, this allows a bias to be applied to the photoconductivegap (as shown in Fig. 2), or for receivers the current flow canbe measured. The electrodes and photoconductive gap aregenerically termed the photomixer; the element responsiblefor modulating the response of the electrical circuit accordingto the optical field. The equivalent circuit for the photomixer,antenna and external circuit is shown in Fig. 3.

While the simple modulation of conductance mechanism maybe clear, the role of the antenna in coupling charge movement tofree-space radiation is less so. Dipole separation and terahertzemission can occur in the absence of an antenna and an appliedbias, as in the observation of surface field terahertz emission[20]. The emitted power can, however, be increased by usingelectrodes to apply an external bias that exceeds that associatedwith the surface depletion field. Nevertheless, even in this case,radiation to free space may still arise from charge movementwithin the photoconductor itself, prior to charge transfer to theelectrodes of the circuit, and radiation from a coupled antenna.

An additional enhancing effect of adding the semiconductor-metal junction may be inferred from the increased terahertzpower observed when carriers are optically injected near to theanode [21]. This “near-anode” enhancement is explained by thehighly nonlinear electric field distribution across the photocon-ductive gap, caused by the formation of a space charge field atthe interface. A similar enhancement is seen when near singularelectric fields are produced using a sharp “bow-tie” geometryfor the electrodes [22]. This is attributed to an enhanced con-tribution from the fields associated with the flux of excited car-riers that reach the interfaces before scattering, and is particu-larly prominent at the anode, since the mobility of the electronsis greater than that of the holes.

The motivation for the work presented here is the systematicoptimization of photomixer and antenna structures for all-op-toelectronic CW-THz imaging systems [19], [23], [24], usingall of the design criteria discussed above. The theory of pho-tomixing, and the equations governing the power transformedto terahertz radiation are presented in Section II, to allow theparameters governing the overall device performance to be de-termined. Sections III–V then summarize and address the re-

GREGORY et al.: OPTIMIZATION OF PHOTOMIXERS AND ANTENNAS FOR CW TERAHERTZ EMISSION 719

quirements imposed upon the photoconductor material, elec-trodes and antenna design, respectively. The electrode designis assessed using terahertz maps to deduce the contribution tothe terahertz emission from illumination of different parts of thephotomixer structure. These maps are used to determine the op-timum electrode geometry, using the near-anode enhancementto improve the efficiency of modulation from the photomixer.Section V details design considerations for the antenna, withnumerical simulations to calculate the frequency response. Sec-tion VI then describes our experimental work to correlate theterahertz polarization with the frequency of emission, allowinginferences to be made regarding the separate role of the photo-conductor and antenna in emission at different frequencies.

II. PHOTOMIXING MECHANISM

For two visible or near-infrared CW laser beams, collinear inspace and with aligned linear polarization, the total electric fieldis the linear sum of the two individual fields, modulated at thedifference frequency. Owing to the finite excitation and decaytimes for the photo-created carriers, a semiconductor is able torespond only to the slower difference frequency beats, allowingan oscillation in the conductance to be produced at terahertz fre-quencies. Thus terms that vary at optical frequencies, , or

, may be replaced by their mean (time-averaged) values.For two beams with powers and , and frequencies and

, the combined instantaneous power incident upon the semi-conductor is then

(1)

where , and is introduced as a parameter todescribe the quality of the spatial overlap, and varies between 0(no overlap) and 1 (perfectly matched) [15].

The carrier density in the photoconductive gap may be de-scribed by

(2)

where is the instantaneous number density of electron–holepairs, is the external quantum efficiency (number of photocre-ated carrier pairs per incident photon), is the active area (pre-sumed equal to the laser spot area), and is the absorption depth.

is the mean photon energy of the visible/near-infrared laserbeam. The second term in (2) gives the exponential decay of car-riers from the conduction band with a decay lifetime char-acterized by . Substituting (1) into (2) gives a differential equa-tion with the analytical solution

(3)Factors of the form also appear, but are neglectedsince they saturate on picosecond timescales. Ignoring the resis-tance of the contacts, the time-dependent conductance,of the photoconductive gap may be written as

(4)

where is the bulk conductivity, is the effective carrier mo-bility, and is the width of the photoconductive gap. Thus to-gether, (3) and (4) give an estimate of the time-dependent con-ductance of the photoconductive gap.

Using the equivalent electrical circuit shown in (Fig. 3), thetotal impedance is given by

(5)

When a dc bias of is applied to the external circuit, the in-stantaneous power dissipated in the load (antenna) resistance is

(6)

Substituting (3)–(5) into (6), taking the real part, and using thefact that , gives a closed form expression for thetotal instantaneous power dissipated in the antenna. This con-tains both terahertz (ac) and Joule heating (dc) components

(7)

The mean power of the terahertz component may be found bydisregarding the constant offset and averaging the sinusoidalterms, to give an expression consistent with that derived byBrown et al. [15]

(8)The Joule heating may be found by solving (2) for the steady

state solution (using a constant, time averaged value of ). Thusthe energy dissipated in the photoconductor resistance is givenby

(9)

In practice, the applied bias is governed by the current thresholdat which the power dissipated as heat in (9) causes damage to thedevice. Thus, back substituting from (9) for the bias in (8) givesthe maximum emitted terahertz power at the dc power limit forthe photoconductor. This is equivalent to running the device ata constant dc current compliance

(10)

Equation (10) has the following properties. The output poweravailable is maximized for (perfect overlap), and

, when the total incident laser power is kept constant. At highfrequency, , and thus a roll-off of 12 dB per oc-tave is expected. Differentiation of (10) with respect to yieldsthe condition for the optimum carrier lifetime in the material

(11)

720 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 41, NO. 5, MAY 2005

Fig. 4. Plots of the theoretical maximum terahertz power output as a functionof the semiconductor trapping lifetime. As the frequency increases, theoptimum lifetime decreases. Inset is the optimized power plotted as a functionof frequency.

The lifetime dependence of the emitted terahertz power isplotted in Fig. 4 for a variety of beat frequencies. At anyparticular beat frequency, there exists a maxima as given by(11). If the carrier lifetime is too short, the carrier density neverbecomes large enough to produce a significant modulation inthe conductivity. However, if the carrier lifetime becomes muchlonger than the beat period, then the accumulation of chargeincreases the heating in the device, which requires the biasto be decreased in order to avoid damage. Together with thescreening effect of trapped carriers, this dramatically reducesthe induced terahertz current, and hence the power available.Note that even under optimum conditions for each frequency,the maximum available terahertz power decreases by 3 dB peroctave (see inset to Fig. 4). This is intrinsically a result of thegeneration process, which gives proportionally more energydissipated as heat at higher frequencies.

To summarize, (10) predicts the conditions required to en-able the power at any given frequency to be maximized. Theperformance of photomixing devices as a function of frequencyis dictated by three effects. First, the finite time required for ex-cited carriers to recombine greatly limits the performance asthe frequency is increased. Second, the finite capacitance of thephotoconductive gap allows an accumulation of charge at theelectrodes. Again, this has the effect of reducing any modula-tion with frequencies greater than the corresponding timeconstant. Finally, the efficiency with which terahertz power iscoupled out of the photomixer and radiated into free space will,in general, be frequency dependent, given by .

It is convenient to view the components associated with thesethree criteria independently, such that a modular approach maybe taken to optimization. First, the carrier lifetime is optimizedby use of low-temperature-grown gallium arsenide (LT-GaAs)grown and annealed under controlled conditions (Section III).The second design element is the optimization of the electrodegeometry to control the capacitance, while simultaneously uti-lizing near-anode enhancements (Section IV). Finally, the effi-ciency of free-space coupling is optimized as a function of theimpedance match between the photomixer and antenna (Section

V). The mean impedance of the photoconductive gap may be es-timated using (4), and putting typical values of cm ,

m, cm , m , andm predicts the impedance of the photoconductive gap to be

large 10 k . Hence, the impedance of the antenna must bemaximized at the desired frequency: resonant structures will beshown to increase power output by up to an order of magnitudeat the target frequency when compared to conventional broad-band antennas [25].

III. MATERIAL AND DEVICE FABRICATION

For CW mixing in the frequency range 0.2–2 THz, optimumphotocarrier lifetimes range from 80 to 800 fs. GaAs grown bymolecular-beam epitaxy (MBE) has electron–hole recombina-tion lifetimes of typically 10 ps–1 ns. The effective conductionband lifetime can be significantly reduced, however, through theintroduction of mid bandgap traps. These might be associatedwith the inclusion of point defects in the GaAs lattice, and cre-ated, for example, either by post-growth ion-implantation, or byLT-GaAs [8], [26]–[28], as used in this work. Such defects canarise from an excess of arsenic, and they provide a large cap-ture cross section for the trapping of conduction band electrons,leading to trapping times as short as 90 fs [29]. Subsequently, thematerial may be annealed to increase the resistivity, albeit witha corresponding increase in lifetime [30]. This compromise be-tween short carrier lifetimes and high resistivities depends crit-ically upon the precise growth and anneal conditions.

In fact, it has recently been shown that the optimum annealtemperatures lie well below those commonly reported, with ap-preciable beneficial effects occurring for temperatures as low as300 C [31]. This enables the production of photoconductivedevices with high resistivities, and essentially no compromise inthe carrier lifetimes, and is illustrated in Fig. 5, where the carrierlifetime and resistance of LT-GaAs is plotted as a function ofanneal temperature. The lifetime was measured using time-re-solved photoreflectance [28], [32]–[34] using 90-fs pulses froma Ti:sapphire laser operating at 800 nm (inset). The accuracyof such measurements is dependent upon the wavelength andpower of the exciting laser, as well as the correct interpretationof various artifacts that can arise. In addition, photoreflectancecurves take no account of the possible effect of the electric fieldon the lifetime [35], [36]. However, the technique provides valu-able and quantitative estimates for the lifetime, which are suf-ficient for comparative measurements. The resistivity was char-acterized through measurement of the resistance of a test device,with a 5- m photoconductive gap, as pictured inset in Fig. 9.

The excess arsenic incorporation during low-temperaturegrowth results primarily in the formation of antisite defects(arsenic atoms located at gallium sites) and gallium vacancies.Annealing then provides the thermal activation energy, whichallows individual defects to migrate through the material toform metallic precipitates. At any given annealing tempera-ture, these defect concentrations will approach an equilibriumvalue as the duration of the anneal increases, according to thevacancy assisted diffusion mechanism applied by Bliss et al.[37]. The position of this equilibrium is a sensitive function ofthe anneal temperature: for temperatures in excess of 550 C

GREGORY et al.: OPTIMIZATION OF PHOTOMIXERS AND ANTENNAS FOR CW TERAHERTZ EMISSION 721

Fig. 5. The carrier lifetime and resistance of a photoconductive test device asa function of anneal temperature [31]. Material annealed in region I shows asubstantial increase in resistance for no compromise in lifetime. In region II,the resistance increases by two orders of magnitude, for a small increase in thelifetime. Further increase of the temperature dramatically increases the lifetime,with little further increase in resistivity. Inset is a sample reflectivity curve for a375 C anneal.

though, there is essentially complete elimination of the pointdefects. Even at the lowest anneal temperatures, any change inthe arsenic content is also accompanied by an increase in theresistivity of the material, and anneal temperatures between350 C and 550 C (regions I and II in Fig. 5) were found togive the optimum combination of lifetime and resistance forCW photomixing.

For the photoconductive material in this work, LT-GaAs wasdeposited by MBE onto an undoped, semi-insulating (001)GaAs substrate. The wafers incorporated an epitaxial bufferlayer of GaAs grown at 600 C followed by a 100-nm insulatingAlAs to prevent parallel conduction through the substrate.This was followed by a 1- m thick epilayer of GaAs grown at200 C 10 , with the temperature estimated using a ther-mocouple. Following growth, the wafers were annealed ex situin a rapid thermal annealer under a nitrogen atmosphere, andwith a semi-insulating GaAs wafer being used to passivate thesurface and inhibit arsenic loss. The anneal was performed at atemperature of 475 C for 10 min, producing a carrier lifetimeof 200 fs, and an estimated bulk resistivity of 10 cm.

The electrode structure and antenna fine-features were de-fined in polymethyl methacrylate (PMMA) resist using elec-tron beam lithography, with lateral dimensions down to 200nm. The macroscopic antenna structure was then superimposedusing optical lithography in a separate exposure. Both the pho-tomixer electrodes and antennas consist of 20–20–400 nm ofTi–Pd–Au—the large thickness is necessary to aid heat conduc-tion away from the photoconductor, and to allow large photocur-rents ( 1 mA) to flow without damage to the device.

IV. PHOTOMIXER CHARACTERIZATION

Previous studies of the optimum geometry for photomixershave investigated the distribution of the electric field in the gap,both laterally and as a function of depth [38]. It was found thatthe change in the trap occupation of the photoconductor closeto metal-semiconductor junctions gives a highly nonuniformelectric field distribution, leading to effects such as near-anode

enhancement [21]. The electric field away from the electrodesis then much weaker, contributing little to the terahertz power.In fact, measurements have shown that almost 90% of thepotential difference is dropped within a few microns of theanode [21]. For CW-THz photomixers, where the incidentoptical power is relatively low, it is advantageous to increasethe active area for generation by adding interdigitated fingersto the electrodes. This effectively increases the length of themetal–semiconductor interface at the electrodes. It also in-creases the conductance and hence improves the impedancematch with the antenna, but at the cost of increased deviceheating.

However, the inclusion of interdigitated photomixing ele-ments also leads to an increased capacitance [16], which scalesapproximately with the active area in which photocarriers areexcited. The area can be reduced to match the diffraction lim-ited laser beam, but there are penalties in terms of the currentlimit for device damage, and resilience of the system to anydrift in the laser beam alignment. Compromise is also necessaryin the choice of dimensions for the interdigitated fingers andgaps. The metal fingers must be large enough to define usinglithographic processes and sufficiently robust to withstand highphotocurrents. However, if the width becomes too great, thephotomixer becomes inefficient as much of the semiconductoris obscured. In the patterns assessed in this work, the activeareas were varied from m m up to m msquare, with the number of fingers in the pattern changed from3 to 11, maintaining a finger separation of 1.7 m.

To investigate the geometrical origin of the terahertz emis-sion, a terahertz map of the photomixer was constructed by plot-ting the terahertz output as a function of the laser spot positionon the active area. This experiment was performed using time-domain (pulsed) terahertz apparatus [12], with a femtosecond(120-fs pulse duration, 75-MHz repetition rate) laser being usedto excite broadband terahertz pulses in the photomixer. Detec-tion was fully optoelectronic, using a component separated fromthe original laser beam to gate the receiver. A rapid-scan delayline was used to collect the detected terahertz waveforms at arate of 15 Hz. In each case, a standard 60 “bow-tie” antennaand 25-mm-diameter hyper-hemispherical silicon lenses wereused to couple the radiation in and out of the semiconductors.Overviews of the emission and detection processes are givenin [3].

To obtain the maps, the tightly focused laser spot 1.7 mwas raster scanned across the photomixer in both and di-rections. Thus carriers were created only in local areas of thefinger pattern. The amplitude of the terahertz pulse is plottedas a function of position to produce an emission map of thephotomixer. An example for the test device with a 5- m gapis shown in Fig. 6, where the terahertz power is plotted for il-lumination along a straight line running from one electrode tothe other, across the center of the gap, as illustrated inset. Adiagram of the laser response function (assumed gaussian) isalso included, and used to deconvolute the data, as shown by thedashed curve. The electrodes (shown shaded) cause no terahertzoutput when illuminated since no photoconduction is occurring.The near-anode enhancement is clear, with the terahertz powerdropping by 58% within 1 m of the electrodes. This effect is

722 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 41, NO. 5, MAY 2005

Fig. 6. Plot of terahertz-emitted power as a function of the laser excitationposition in a 5-�m photoconductive gap (inset). The dashed curve shows thesame data deconvoluted using the Gaussian laser response function (inset).The electric field profile may be deduced from this—over half of the potentialdifference occurs within 1 �m of the electrodes.

Fig. 7. (a) Image to show a 2-D terahertz map of the output power as afunction of the position of the exciting laser spot. The areas for which theincident laser spot produces the least terahertz amplitude are shown in black,ranging to the largest detected amplitude shown in white. (b) Photograph of thestructure, for comparison. Each image corresponds to a 30-�m square regionof the photomixer. (c) Line-scan across the center of the device to show theconvoluted terahertz power (and thus electric field profile) for a single y value,as shown dashed in the micrograph.

seen near to both electrodes, since the bias was alternated to fa-cilitate lock-in detection.

Thus the design principle for the optimization of the fingerseparation in an interdigitated device is to maximize the field inthe photoconductive gap by allowing the two enhanced regionsto overlap. Fig. 7(a) shows the two-dimensional (2-D) terahertzmap generated by a finger pattern with an active area of m

m and with 11 interdigitated fingers each of width 300 nm, a

micrograph showing the same area of the emitter is also included[Fig. 7(b)].

Fig. 7(c) shows the raw (convoluted) data for the terahertzpower profile extracted along a line of constant , shown in themicrograph. Deconvoluting the response curve of the laser beam(not shown) yields an electric field profile with no sharp fea-tures, indicating that the field (and terahertz emission) is fairlyconstant across the device. Thus the minima in the raw data re-flect only the loss of optical power owing to the laser beam beingobscured by the electrodes. An optimum occurs when the lateraldistance over which the electric field contributes to the terahertzpower is comparable to both the finger spacing and the laserspot diameter (1–2 m). Perhaps surprisingly, there is no sig-nificant enhancement in output at the fingertips, where electricfield singularities might be assumed to increase the power sub-stantially. All of these observations are in agreement with Cai etal. [22], who stated that the near-anode enhancement is absentin the small gap limit, without detailed explanation.

A pulsed terahertz system was used here as an assessmenttool, because of its high signal-to-noise ratio (SNR) and abilityto provide information for all frequencies. Although it is ac-cepted that the generation mechanism of pulsed and CW-THzemission are subtly different, it is expected that a strong corre-lation will exist, since both will depend critically on the electricfield profiles in and around the interdigitated finger patterns.

The photomixer designs assessed were next tested asCW-THz emitters, with excitation produced by two, inde-pendent, external-cavity diode lasers (TOPTICA PhotonicsDL100) operating at wavelengths around 850 nm, with a nom-inal instantaneous linewidth of order 1 MHz. The wavelengthof each laser can be tuned by several nm, giving a beat fre-quency range of approximately 10 GHz–3 THz, with a 4 GHzmode spacing (resolution). The beams were made collinearusing a 50:50 beamsplitter, giving a 40-mW combined beam,which was focused onto the photomixer elements using a 10microscope objective. The terahertz radiation was coupledout of the rear surface of the substrate using a high-resistivitysilicon lens, and detected with a silicon composite bolometer.The applied dc bias of up to 30 V was modulated at 200 Hzfor lock-in detection. To prevent damage to the structure fromJoule heating, the bias was limited by a dc current complianceof 0.75 mA. This corresponds to a total dissipated electricalpower of approximately 25 mW—comparable to the incidentoptical power.

The emission power is plotted in Fig. 8 as a function of thelaser difference frequency for photomixers with different num-bers of fingers, but with a constant separation of the optimum1.7 m as characterized previously (micrographs shown inset,to scale). In each case, the photomixer is loading a broadbandspiral antenna, consisting of self-complementary 4-turn spirals,with inner and outer final radii of 1.9 and 2.3 mm, respectively(shown inset to Fig. 10). The broad water vapor absorption bandabove 1 THz is clear in all cases, with the well-known spectrallines at 1.10, 1.16, and 1.41 THz also in evidence, as indicatedby the arrows.

The performance of each design is broadly equivalent in thefrequency range of 0.1–0.4 THz. Beyond this, the capacitiveroll-off in output occurs at a higher frequency as the number

GREGORY et al.: OPTIMIZATION OF PHOTOMIXERS AND ANTENNAS FOR CW TERAHERTZ EMISSION 723

Fig. 8. Plot of the emitted power as a function of frequency for three of thedevices tested. The inset micrographs (all at the same scale) show the geometryof the interdigitated-finger photomixers. At higher frequency, the power is seento roll off more rapidly as the device capacitance increases. The dashed lineshows the slope corresponding to a rolloff of �12 dB per octave, as predictedtheoretically.

of fingers and the active area is reduced. The decrease at highfrequency ( 1 THz) is characterized by 12 dB per octave, de-rived in (10). This difference amounts to more than an order ofmagnitude at 1.4 THz for devices A and C. Using an electromag-netic simulator in the frequency range 0–3 THz, numerical cal-culations predict the total capacitances to be approximately 2.8and 18 fF, respectively. These are consistently greater than thevalues analytically calculated for similar structures by Brown etal. [39], owing to the inclusion of screening and the increasedmetallization thickness in our case. A photomixer capacitanceof 2.8 fF, together with a 72- spiral antenna, yields antime constant of 200 fs—a value comparable to the optimizedphotocarrier lifetime in LT-GaAs.

To summarize, the optimum finger spacing appearsto lie close to the dimensions chosen by Brown et al. [38](0.2- m fingers separated by 1.8- m gaps), and is independentof frequency. The optimum choice of total active area, incontrast, very much depends upon the working frequency ofthe photomixer. Equation (10) implies that the width of thephotoconductive gap should be minimized to optimize theemitted power. However, this relationship holds only as longas the active area can be matched to the laser spot size, andis able to dissipate heat from the optical power. At higherfrequencies, the need for a low capacitance dominates, toensure that , requiring small active areas. At lowerfrequencies, however, the capacitance of the device is largelyirrelevant. In fact, it may be preferable in this instance to usea larger active area. This would permit higher current flowswithin the device [and increased from (10)], boostingthe power from device C in Fig. 8, for which constant currentswere employed. Furthermore, a larger device area will be lesssensitive to beam drift from the CW lasers.

V. ANTENNA CHARACTERIZATION

The antenna transforms the terahertz current in the pho-tomixer into an electromagnetic wave that can propagate in freespace. The characteristic impedance of free space is 120 ,

Fig. 9. Simulations of two bow-tie antennas with 5-�m photoconductivegaps (micrographs of the actual structures shown inset). (a) With a barephotoconductive gap. (b) With interdigitated fingers. For both structures,the antenna radiation impedance is plotted as a function of frequency. Thecapacitive effect of the fingers causes the effective impedance (and henceradiated power) to fall sharply as the frequency is increased. The fingers weremodeled as a lumped capacitance of 1 fF; the oscillations at low and highfrequencies are numerical artifacts.

independent of frequency. However, the real part of the pho-tomixer output impedance as seen by the antenna is frequencydependent and much higher (of order 10 k ) [25]. Thus in prac-tice, impedance matching is difficult, and the radiated powerscales with the input impedance presented by the antenna.For pulsed terahertz systems, a broad range of frequencies(typically between 50 GHz and 2 THz) are simultaneouslygenerated. In this case, the antenna must be efficient over awide frequency range, leading to the widespread use of bow-tieantenna designs. These are intrinsically broadband and easyto fabricate. An alternative type of broadband antenna designis the self-complementary log-spiral with a central feed [18].The radiation impedance of this design is largely frequencyindependent, with a nominal calculated value of 72 (for anyself-complementary geometry), in a GaAs dielectric half space[16].

For our antenna designs, we calculated numerically the an-tenna input impedance as a function of frequency, using com-mercially available electromagnetic simulation software [40].The substrate–air interface and antenna geometry were definedwithin a finite-element mesh and metal electrodes added as per-fect electrical conductors. The photomixer was modeled by gen-erating a gaussian wave packet in the gap and examining thefrequency-dependent complex current and voltage at the an-tenna feed. The effect of including interdigitated fingers at thefeed-point was modeled by a lumped capacitance in parallelwith the photoconductive active area. The capacitance was cal-culated numerically for each geometry using the same software.

The calculated response is shown in Fig. 9, for a truncated60 bow-tie antenna with a 5- m photoconductive gap [inset(a) to Fig. 9]. It may be immediately seen that the bare antennadoes have a relatively flat response (dotted line), dropping from90 at low frequencies to about 70 at 2 THz, as expected.When an interdigitated photomixer feed [inset (b) to Fig. 9], isadded to the antenna, the additional capacitance causes the netimpedance presented to the photomixer to fall dramatically at

724 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 41, NO. 5, MAY 2005

frequencies beyond 1 THz. The decrease in power is an order ofmagnitude at 2 THz.

For coupling of terahertz standing wave oscillations from theantenna, it is necessary for the antenna to have features on theeffective length scale of the emitted terahertz wavelengths. Thusbroadband antennas are designed with a fractal geometry: boththe bow-tie and the spiral antenna designs fall into this cate-gory. However, for CW-THz systems, intended for operation ata single frequency, it is often desirable to maximize the inputimpedance at a single, target frequency. Modeling and experi-mental work has begun to address the use of resonant dipole an-tennas for this purpose [41]. However, only recently have suchantennas been optimized to enhance the monochromatic emis-sion from photomixers [25]. Duffy et al. used this concept toobtain higher frequency operation (beyond 2 THz). This wasachieved by using the inductance present in dual dipole an-tennas to tune out the capacitance, thus increasing the frequencythreshold for capacitive roll-off. Resonances at frequencies upto 2.7 THz were achieved, with the peak power subject only tothe 6 dB per octave roll-off owing to the carrier lifetime. Forour work, in contrast, we sought to design low-frequency ( 1THz) resonant antennas that increase power output in emission,and SNR in detection.

The resonant behavior of a center-feed dipole antenna(300 m long, 8 m wide, and of 1 m thickness) wasmodeled numerically, and predicted a clear impedance reso-nance, shown in Fig. 10. The magnitude of the resonant poweroutput was increased substantially by the use of quarter-wavechoke elements to prevent radiative power loss along the biastransmission lines. Although previous work also included theeffect of ohmic losses in the conductors [25], in our case thereis no significant effect on the value of the resonant frequency, orthe optimization condition. A decrease in the power output and

-factor would be expected: but the effect is less severe becausethe dimensions of the conductors are increased (10–20 m inwidth and 440 nm in thickness) relative to the 1 m conductorsin [25].

This antenna was then fabricated, with the dipole antenna,photomixer and choked feeds being defined lithographically onannealed LT-GaAs with a sub-200-fs lifetime. Using a CW-THzsystem, the emitted power was then measured as a function ofthe beat frequency, using a silicon composite bolometer, andFig. 10 shows the predicted bolometer signal plotted as a func-tion of frequency for the choked dipole antenna, together withexperimental data for both (a) choked dipole and (b) spiral an-tennas.

The simulation predicted a resonance at 0.41 THz witha peak drive-point impedance of 270 , and a full-width athalf-maximum (FWHM) of 40 GHz. Experimentally, excellentagreement was seen, with a resonance occurring at 0.40 THzwith a FWHM of 80 GHz. The resonance increases the poweroutput by over 6.5 dB, compared to that achieved with thebroadband spiral antenna fabricated on the same material, andexcited in an identical way. The width of the resonance appearsgreater than predicted by the model, and may be associated withohmic losses in the metallic structures. However, the experi-mental resonance is still much narrower than seen previouslyfor this type of antenna: the responses reported in [2], [25]

Fig. 10. Simulated and experimental results from (a) 300-�m dipole and (b)broadband spiral. The simulation data (curve) has been normalized with respectto the experimental results at resonance.

typically exceeded 200 GHz in width, at central frequenciesof 0.7–1.0 THz. The reason for this difference is not clear atpresent.

The resonant antenna was then used as a receiver, in placeof the bolometer, and gated using the beat frequency of thelasers, with a total incident power of 20 mW. The relativephase between the incident terahertz radiation and the beatenvelope was changed using a delay line, generating an interfer-ogram. The Fourier transform is then the convoluted frequencyresponse of both emitter and detector. Initially, a resonantantenna and broadband (spiral) receiver were characterized forreference. When a resonant receiver matched to the emitterwas substituted, the detected signal amplitude increased by 9.8dB at 0.4 THz. An added benefit of this approach is that thedetector now also acts as a filter, and thus any RF componentsare largely rejected.

VI. POLARIZATION MEASUREMENTS

The polarization of the emitted terahertz radiation can be re-lated to the geometry of the antenna. A bolometer is essentiallyinsensitive to both frequency and polarization, and so to mea-sure the polarization experimentally, a wire-grid terahertz polar-izer was inserted into the terahertz path. Detected powers weremeasured as a function of both frequency and polarizer anglefor spiral, bow-tie and dipole antennas, across the entire mea-surable frequency range with the angle between the polarizingfilter elements and the long axis of the interdigitated fingersvaried between 0 and 180 . The ratio between the minimumand maximum powers transmitted through the polarizer may be

GREGORY et al.: OPTIMIZATION OF PHOTOMIXERS AND ANTENNAS FOR CW TERAHERTZ EMISSION 725

Fig. 11. Extinction ratio (minor/major) of the elliptically polarized emittedterahertz as a function of frequency. A value of 0 corresponds to a perfectlylinear polarization, whereas a value of 1 indicates a perfectly circularpolarization state.

used to define the extinction ratio. This is plotted in Fig. 11 foran 8 m 8 m photomixer coupled to both a broadband spiral(dotted line) and simple bow tie antenna (solid line).

Fig. 11 shows that the bow-tie antenna emission is linearlypolarized over the entire frequency range, with an extinctionratio of order 1:20. A similar experiment was performed fora resonant dipole antenna, and the linear polarization state atresonance had an extinction ratio of better than 1:650. Thisis expected because the electric field is directed along thedipole length. The emission from the spiral, however, is almostcircularly polarized (orthogonal components equal) at 0.1 THz,but this state becomes increasingly elliptically polarized as thefrequency rises. Above 0.6 THz, the emission is essentiallylinear, as for the bow tie antenna. The change in the polarizationstate with frequency in the spiral antenna may be attributedto the change in the antenna geometry as seen by differentwavelengths. Wavelengths that are resonant with the spiralarms will be emitted with a circular polarization.

To illustrate this further, Fig. 12 shows the terahertz poweron a polar plot, as a function of the angle between the interdigi-tated fingers and the transmission axis of the wire grid polarizer.It may be seen that as the emitted terahertz radiation becomesmore elliptically polarized with increasing frequency, the orien-tation of the major axis of polarization also rotates, followingthe direction of the spiral antenna arms. As the linear limit isapproached at the center of the spiral, this major axis becomesoriented parallel to the interdigitated fingers on the photomixer.Thus we deduce that in this regime, the emission is radiated di-rectly from linear dipole resonances in the interdigitated fingers.The linearly polarized output from the bow-tie (and dipole) an-tennas is always in the direction parallel to the fingers over themeasurable frequency range.

Polarization measurements were then repeated with similarantenna structures, using a pulsed terahertz system based on thefemtosecond laser. The improved SNR allowed an assessmentat much higher frequency, and with the advantage that the po-larization may be measured for all frequencies simultaneously.A pair of time domain traces with the terahertz grid polarizer

Fig. 12. Polar plots of the terahertz radiation emitted from the spiral antennaat frequencies of 0.10, 0.16, 0.23, 0.37, 0.48, and 0.60 THz. 0 and 90correspond to the components of terahertz power with the electric fieldperpendicular and parallel to the fingers, respectively.

Fig. 13. Extinction ratio, derived from the Fourier transforms of orthogonalcomponents, plotted as a function of frequency. The bow tie antenna withoutinterdigitated fingers (dotted line) has a linear polarization across the wholefrequency range. The spiral antenna with fingers (solid line) has a morecomplicated dependence, owing to the spiral structure at low frequency (below0.6 THz) and to the fingers at high frequency (above 1.1 THz). Thus thedependence may be divided into three regimes, where the effective wavelengthis on the scale of (a) the spiral arms, (b) the finger length, and (c) the fingerseparation.

in orthogonal orientations were obtained. By dividing the cor-responding Fourier transforms, the extinction ratio as a functionof frequency was deduced. Fig. 13 shows the frequency depen-dence of the extinction ratio, defined once again, as the ratio ofthe power emitted with the terahertz electric field perpendicularto the fingers, to the power emitted with the electric field orien-tated parallel to the fingers. Thus a value of 0 corresponds to theterahertz linear polarization being entirely parallel to the fingers,

726 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 41, NO. 5, MAY 2005

while a value of 1 describes a situation where the componentsin the orthogonal directions are equal.

The dashed line in Fig. 13 shows results for a bow-tie an-tenna with no photomixer fingers. As with the CW measure-ments, a linear polarization (more than 95% linear) is observedacross the entire frequency range, with the major axis alignedparallel to the bias field in the photoconductive gap. The solidline shows the analogous curve for a spiral antenna with inter-digitated fingers defined at the center, shown inset on variousscales. A value of unity is measured for the extinction ratio atlow frequency, indicating circular polarization below 0.2 THz,where the effective wavelength is on the scale of the spiral arms[inset (a)]. As with the CW measurements, the polarization is el-liptical between 0.2 and 0.6 THz. The peak at 0.4 THz is repro-ducible, and may be due to a resonance at the finger-spiral inter-face. This may be an artifact of the processing technique, sincethe fingers and spiral were necessarily defined in two separatemetallization processes. Between 0.6 and 1.1 THz, the polariza-tion is linear, and aligned parallel with the fingers. Again, thisis in agreement with the CW measurements, and suggests thatthe emission in this frequency range arises from a resonance onthe length scale of the fingers [inset (b)]. At frequencies above1.3 THz, corresponding to length scales substantially below theresonance lengths of the fingers, a component in the perpendic-ular direction (i.e., directly radiated from current transients inthe semiconductor material between the fingers) becomes ob-servable [Fig. 13(c)]. This causes the extinction ratio to riserapidly with increasing frequency for the emitter with fingers.The similarity of the polarization behavior with increasing fre-quency for both pulsed and CW-THz sources justifies the asser-tion that the role of the antenna is identical in each case, and dis-tinct from the photomixer, where differences may occur owingto the vast different in peak carrier densities between the twocases.

The idea of a characteristic length scale is central to the theoryof dipole antennas in radio or microwave applications. The anal-ogous theory for planar terahertz transceivers takes into accounta modification owing to the dielectric half-space of GaAs

. For the fundamental (half-wave) dipole resonance, the ef-fective resonant length scale is given by [42]

(12)

However, experiments using full- and half-wave dipoles [41]show that these simple arguments are incomplete, whereas ourpolarization measurements yield direct estimations of theselength scales. The polarization state of the emission from thespiral antenna changes from elliptical to linear at a frequency ofapproximately 0.6 THz, where the calculated half-wavelengthsin free space and the GaAs half-space are 250 and 95 m,respectively. In contrast, the corresponding physical lengthscale at the spiral center where this polarization change occursis only 20–60 m; such discrepancies occur since (12) holdsonly for an ideal isolated, thin dipole. In fact, the resonantbehavior of real devices is strongly perturbed by both the finitestructure width and thickness, as well as the electrical responseof the adjoining bond wires—a conclusion supported by ournumerical modeling.

VII. CONCLUSION

We have outlined a systematic approach for the improvementof CW emitters and receivers that operate on the principle ofphotomixing. The photoconductor, electrode geometry and an-tenna have each been independently optimized according to theindividual criteria from our theoretical model.

There are advantages in using CW-THz radiation for com-mercial imaging and spectroscopy systems, compared to themore established pulsed technologies. The narrow bandwidthyields a very high spectral density, allowing improved frequencyresolution and faster scanning times [24]. Furthermore, use ofinexpensive diode lasers in a portable terahertz system will becompact, robust, genuinely turn-key and of low cost [19]. How-ever, diode lasers provide relatively little power, and thus in-creasing the efficiency of both the photomixers and antennas arekey steps toward achieving their practical implementation.

REFERENCES

[1] D. H. Auston, “Picosecond optoelectronic switching and gating in sil-icon,” Appl. Phys. Lett., vol. 26, no. 3, pp. 101–103, 1975.

[2] P. R. Smith, D. H. Auston, and M. C. Nuss, “Subpicosecond photocon-ducting dipole antennas,” IEEE J. Quantum Electron., vol. 24, no. 2, pp.255–260, Feb. 1988.

[3] P. U. Jepsen, R. H. Jacobsen, and S. R. Keiding, “Generation and de-tection of terahertz pulses from biased semiconductor antennas,” J. Opt.Soc. Amer. B., vol. 13, no. 11, pp. 2424–2436, 1996.

[4] R. Köhler, A. Tredicucci, F. Beltram, H. E. Beere, E. H. Linfield, A. G.Davies, D. A. Richie, R. C. Iotti, and F. Rossi, “Terahertz semiconductor-heterostructure laser,” Nature, vol. 417, pp. 156–159, 2002.

[5] M. C. Kemp, P. F. Taday, B. E. Cole, J. A. Cluff, A. J. Fitzgerald, andW. R. Tribe, “Security applications of terahertz technology,” Proc SPIE,vol. 5070, pp. 44–52, 2003.

[6] D. D. Arnone, C. M. Ciesla, and M. Pepper, “Terahertz imaging comesinto view,” Phys. World, vol. 13, no. 4, pp. 35–40, 2000.

[7] J. V. Rudd and D. M. Mittleman, “Influence of substrate-lens design interahertz time-domain spectroscopy,” J. Opt. Soc. Amer. B, vol. 19, no.2, pp. 319–329, 2002.

[8] F. W. Smith, H. Q. Le, V. Diadiuk, M. A. Hollis, A. R. Calawa, S.Gupta, M. Frankel, D. R. Dykaar, G. A. Mourou, and T. Y. Hsiang, “Pi-cosec-ond GaAs-based photoconductive optoelectronic detectors,” Appl.Phys. Lett., vol. 54, no. 10, pp. 890–892, 1989.

[9] Y. Chen, S. Williamson, T. Brock, F. W. Smith, and A. R. Calawa, “375-GHz-bandwidth photoconductive detector,” Appl. Phys. Lett., vol. 59,no. 16, pp. 1984–1986, 1991.

[10] M. Tani, K. Sakai, and H. Mimura, “Ultrafast photoconductive detectorsbased on semi-insulating GaAs and InP,” Jpn. J. Appl. Phys., vol. 36, no.2, pp. L1175–L1178, 1997.

[11] S. Kono, M. Tani, P. Gu, and K. Sakai, “Detection of up to 20 THz witha low-temperature-grown GaAs photoconductive antenna gated with 15fs light pulses,” Appl. Phys. Lett., vol. 77, no. 25, pp. 4104–4106, 2000.

[12] C. Baker, I. S. Gregory, W. R. Tribe, I. V. Bradley, M. J. Evans, M.Withers, P. F. Taday, V. P. Wallace, E. H. Linfield, A. G. Davies, andM. Missous, “Terahertz pulsed imaging with 1.06 �m laser excitation,”Appl. Phys. Lett., vol. 83, no. 20, pp. 4113–4115, 2003.

[13] B. B. Hu and M. C. Nuss, “Imaging with terahertz waves,” Opt. Lett.,vol. 20, no. 16, pp. 1716–1718, 1995.

[14] D. M. Mittleman, R. H. Jacobsen, and M. C. Nuss, “T-ray imaging,”IEEE J. Sel. Topics Quantum Electron., vol. 2, no. 3, pp. 679–692,May–Jun. 1996.

[15] E. R. Brown, F. W. Smith, and K. A. McIntosh, “Coherent millimeter-wave generation by heterodyne conversion in low-temperature-grownGaAs photoconductors,” J. Appl. Phys., vol. 73, no. 3, pp. 1480–1484,1993.

[16] E. R. Brown, K. A. McIntosh, K. B. Nichols, and C. L. Dennis, “Pho-tomixing up to 3.8 THz in low-temperature-grown GaAs,” Appl. Phys.Lett., vol. 66, no. 3, pp. 285–287, 1995.

[17] S. Verghese, K. A. McIntosh, S. Calawa, W. F. Dinatale, E. K. Duerr,and K. A. Molvar, “Generation and detection of coherent terahertzwaves using two photomixers,” Appl. Phys. Lett., vol. 73, no. 26, pp.3824–3826, 1998.

GREGORY et al.: OPTIMIZATION OF PHOTOMIXERS AND ANTENNAS FOR CW TERAHERTZ EMISSION 727

[18] S. Verghese, K. A. McIntosh, and E. R. Brown, “Highly-tunable fiber-coupled photomixers with coherent terahertz output power,” IEEE Trans.Microw. Theory Tech., vol. 45, no. 8, pp. 1301–1309, Aug. 1997.

[19] I. S. Gregory, W. R. Tribe, B. E. Cole, C. Baker, M. J. Evans, I. V.Bradley, E. H. Linfield, A. G. Davies, and M. Missous, “Phase sensi-tive continuous-wave THz imaging using diode lasers,” Electron. Lett.,vol. 40, no. 2, pp. 143–145, 2004.

[20] M. B. Johnston, D. M. Whittaker, A. Corchia, A. G. Davies, and E. H.Linfield, “Simulation of terahertz generation at semiconductor surfaces,”Phys. Rev. B, vol. 65, pp. 1653011–1653018, 2002.

[21] S. E. Ralph and D. Grischkowsky, “Trap-enhanced electric fields insemi-insulators: the role of electrical and optical carrier injection,”Appl. Phys. Lett., vol. 59, no. 16, pp. 1972–1974, 1991.

[22] Y. Cai, I. Brener, J. Lopata, J. Wynn, L. Pfeiffer, and J. Federici, “Designand performance of singular electric field terahertz photoconducting an-tennas,” Appl. Phys. Lett., vol. 71, no. 15, pp. 2076–2078, 1997.

[23] T. Kleine-Ostmann, P. Knobloch, M. Koch, S. Hoffman, M. Breede,M. Hofmann, G. Hein, K. Pierz, M. Sperling, and K. Donhuijsen,“Continuous-wave THz imaging,” Electron. Lett., vol. 37, no. 24, pp.1461–1462, 2001.

[24] K. J. Siebert, T. Löffler, H. Quast, M. Thomson, T. Bauer, R. Leonhardt,C. Czasch, and H. G. Roskos, “All-optoelectronic continuous waveTHz imaging for biomedical applications,” Phys. Med. Biol., vol. 47,pp. 3743–3748, 2002.

[25] S. M. Duffy, S. Verghese, K. A. McIntosh, A. Jackson, A. C. Gossard,and S. Matsuura, “Accurate modeling of dual dipole and slot elementsused with photomixers for coherent terahertz output power,” IEEE Trans.Microw. Theory Tech., vol. 49, no. 6, pp. 1032–1038, Jun. 2001.

[26] S. Gupta, M. Y. Frankel, J. A. Valdmanis, J. F. Whitaker, G. A. Mourou,F. W. Smith, and A. R. Calawa, “Subpicosecond carrier lifetime in GaAsgrown by molecular beam epitaxy at low temperatures,” Appl. Phys.Lett., vol. 59, no. 25, pp. 3276–3278, 1991.

[27] S. Gupta, J. F. Whitaker, and G. A. Mourou, “Ultrafast carrier dynamicsin III-V semiconductors grown by molecular-beam epitaxy at very lowsubstrate temperatures,” IEEE J. Quantum Electron., vol. 28, no. 10, pp.2464–2472, Oct. 1992.

[28] H. S. Loka, S. D. Benjamin, and P. W. E. Smith, “Optical characteriza-tion of low-temperature-grown GaAs for ultrafast all-optical switchingdevices,” IEEE J. Quantum Electron., vol. 34, no. 8, pp. 1426–1437,Aug. 1998.

[29] K. A. McIntosh, K. B. Nichols, S. Verghese, and E. R. Brown, “In-vestigation of ultrashort photocarrier relaxation times in low-tempera-ture-grown GaAs,” Appl. Phys. Lett., vol. 70, no. 3, pp. 354–356, 1997.

[30] J. K. Luo, H. Thomas, D. V. Morgan, and D. Westwood, “Transport prop-erties of GaAs layers grown by molecular beam epitaxy at low temper-ature and the effects of annealing,” J. Appl. Phys., vol. 79, no. 7, pp.3622–3629, 1996.

[31] I. S. Gregory, C. Baker, W. R. Tribe, M. J. Evans, H. E. Beere, E. H.Linfield, A. G. Davies, and M. Missous, “High resistivity annealed low-temperature GaAs with 100 fs lifetimes,” Appl. Phys. Lett., vol. 83, no.20, pp. 4199–4201, 2003.

[32] S. Sinha, B. M. Arora, and S. Subramanian, “Photoreflectance andphotoluminescence spectroscopy of low-temperature GaAs grown bymolecular-beam epitaxy,” J. Appl. Phys., vol. 79, no. 1, pp. 427–432,1996.

[33] J.-S. Yu, H.-C. Ho, S. Horng, and C.-C. Chi, “Spectral dependence oftime-resolved photoreflectance of low-temperature-grown GaAs,” Jpn.J. Appl. Phys., vol. 36, no. 4A, pp. 2144–2150, 1997.

[34] M. Stellmacher, J. Nagle, J. F. Lampin, P. Santoro, J. Vaneecloo, and A.Alexandrou, “Dependence of the carrier lifetime on acceptor concen-tration in GaAs grown at low-temperature under different growth andannealing conditions,” J. Appl. Phys., vol. 88, no. 10, pp. 6026–6031,2000.

[35] N. Zamdmer, Q. Hu, K. A. McIntosh, and S. Verghese, “Increase in re-sponse time of low-temperature-grown GaAs photoconductive switchesat high voltage bias,” Appl. Phys. Lett., vol. 75, no. 15, pp. 2313–2315,1999.

[36] K.-G. Gan, J.-W. Shi, Y.-H. Chen, C.-K. Sun, Y.-J. Chiu, and J. E.Bowers, “Ultrahigh power-bandwidth-product performance of low-tem-perature-grown-GaAs based metal-semiconductor-metal traveling wavephotodetectors,” Appl. Phys. Lett., vol. 80, no. 21, pp. 4054–4056, 2002.

[37] D. E. Bliss, W. Walukiewicz, and E. E. Haller, “Annealing of As -re-lated defects in LT-GaAs: the role of gallium vacancies,” J. Electron.Mater., vol. 22, no. 12, pp. 1401–1404, 1993.

[38] E. R. Brown, “A photoconductive model for superior GaAs THz pho-tomixers,” Appl. Phys. Lett., vol. 75, no. 6, pp. 769–771, 1999.

[39] K. A. McIntosh, E. R. Brown, K. B. Nichols, and O. B. McMahon, “Ter-ahertz photomixing with diode lasers in low temperature-grown GaAs,”Appl. Phys. Lett., vol. 67, no. 26, pp. 3844–3846, 1995.

[40] Computer simulation technology. [Online]. Available:http://www.cst.de

[41] K. A. McIntosh, E. R. Brown, K. B. Nichols, O. B. McMahon, W. F.DiNatale, and T. M. Lyszczarz, “Terahertz measurements of resonantplanar antennas coupled to low-temperature-grown GaAs photomixers,”Appl. Phys. Lett., vol. 69, no. 24, pp. 3632–3634, 1996.

[42] M. Tani, M. Watanabe, and K. Sakai, “Photoconductive twin dipole an-tennas for THz transceiver,” Electron. Lett., vol. 38, no. 1, pp. 5–6, 2002.

Ian S. Gregory was born in Stoke-on-Trent, U.K.,in 1979. He received the M.Phys. (hons) degree inphysics from the University of Sheffield, Sheffield,U.K., in 2001, and is currently pursuing the Ph.D. de-gree at the Semiconductor Physics Group, Universityof Cambridge, Cambridge, U.K.

He holds a CASE award with TeraView Ltd., Cam-bridge, U.K., and has now joined their expanding Re-search and Development Group, working on the de-velopment of continuous-wave terahertz sources anddetectors based on photomixing using diode lasers.

Colin Baker was born in Ormskirk, U.K., in 1977.He received the B.Sc. degree in physics from the Uni-versity of Sheffield, Sheffield, U.K. in 2000, and thePh.D. degree in physics from the University of Cam-bridge, Cambridge, U.K., in 2004.

He is currently employed as a member of thesecurity group at TeraView Ltd, Cambridge, U.K.,and his research interests include the applicationof high-power coherent terahertz photoconductivedevices for security screening, and the developmentof cost effective, compact, and commercially viable

terahertz systems.

William R. Tribe was born in London, U.K., in1969. He received the B.Sc. (hons) degree in physicsfrom Imperial College, London University, London,in 1990, and the D.Phil. in the optical propertiesof semiconductors from the Clarendon Laboratory,Oxford University, Oxford, U.K., in 1994.

Following academic posts at Sheffield University,Sheffield, U.K., (1994–1996), and the CavendishLaboratory, Cambridge University, Cambridge,U.K., (1996–2001), studying the electronic andoptical properties of semiconductor materials and

devices, he moved to TeraView Ltd., Cambridge. He now heads an researchand development division responsible for developing novel terahertz systemsand devices including pulsed and continuous-wave systems and is TechnicalLead for applications development in the security area.

Ian V. Bradley was born in Omagh, NorthernIreland, U.K., in 1967. He received the B.Sc. (hons)degree in physics in 1988 and the M.Sc. degree inoptoelectronics and optical signal processing fromQueen University of Belfast, Belfast, U.K., in 1989.He received the Ph.D. degree in interdiffusion ofIII–V semiconductors heterostructures from SurreyUniversity, Surrey, U.K.

Following academic posts at Strathclyde Univer-sity in Glasgow, Glasgow, U.K., (1994–1997), Uni-versity of Nagoya, Nagoya, Japan (1997–1999), and

Felix Free Electron Facility, The Netherlands (1999–2001) studying the opticalproperties of semiconductor materials, he moved to TeraView Ltd., Cambridge,U.K., in 2001. He is currently part of the engineering group within the company.

728 IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 41, NO. 5, MAY 2005

Michael J. Evans was born in Gloucester, U.K., in 1964. He received the B.Sc.degree in materials science from the University of Surrey, Surrey, U.K., in 1988,and the M.Sc. degree in microelectronics from the University of ManchesterInstitute of Science and Technology (UMIST), Manchester, U.K., in 1990.

He is currently employed as a Device Engineer at Teraview Ltd., Cambridge,U.K., and is working on the development of terahertz sources and detectors.

Edmund H. Linfield received the B.A. degree inphysics and the Ph.D. degree in semiconductorphysics from the University of Cambridge, Cam-bridge, U.K., in 1986 and 1991, respectively.

He is currently Professor of Terahertz Electronicsat the University of Leeds, Leeds, U.K. His researchinterests include the development and applicationsof terahertz systems, quantum cascade lasers, molec-ular beam epitaxial growth, and low-dimensionalelectronic systems.

A. Giles Davies received the B.Sc. degree in chem-ical physics from the University of Bristol, Bristol,U.K., in 1987, and the Ph.D. degree in semicon-ductor physics from the University of Cambridge,Cambridge, U.K., in 1991.

He is currently Professor of Electronic and Pho-tonic Engineering at the University of Leeds, Leeds,U.K. His research interests concentrate on the elec-trical and optical properties of low-dimensional elec-tronic systems, multilayered semiconductor devices,and nanotechnology, with particular recent focus on

the development of terahertz frequency systems, and the exploitation of biolog-ical processes for nanoscale assembly.

Mohamed Missous (M’95) received the Ph.D. de-gree from University of Manchester Institute of Sci-ence and Technology (UMIST), Manchester, U.K.,in 1985, focusing on the molecular beam epitaxial(MBE) growth of metals and semiconductors.

He became Chair of Semiconductor Materials andDevices, UMIST, in 2001. All his research has beenclosely associated with the MBE effort at the UMIST,and recently, has concentrated with considerable suc-cess on establishing practical approaches and tech-niques required to meet stringent doping and thick-

ness, control for a variety of advanced quantum devices. This has included workon GaAs, AlGaAs, InGaAs, InAlAs, InGaP, and InAlP. His principal research in-terests are interfaces, both metal–semiconductor and semiconductor–semicon-ductor, and MBE growth mechanisms, especially under conditions of exact sto-ichiometry at low temperatures. He is actively involved as a Consultant to boththe microwave industries in the U.K., Japan, and the U.S. He has published over150 papers on MBE related topics.